Highlights
- Added CUDA Tile backend and profiling support to the Python DSL.
- Expanded compiler frontend and runtime support for kernel argument layouts and launches.
- Improved layout propagation, affine-map diagnostics, iteration-space discovery, and matmul and reduction handling.
- Enhanced
tensor_ir-optwith persistent-kernel options and improved support for large Debug builds. - Expanded MLIR and Python test coverage and refreshed the documentation.
See the README for build and usage instructions.